Graphene field-effect transistors (GFETs) have been used for a variety of applications, including pH sensing, biomarker sensing, and ultra-low noise chemical sensors, among others. In addition to established uses, recent research points to an application in on-chip random number generation (RNG) from an array of GFETs.
RNGs are an important component of a range of industries that are based on technologies such as statistical sampling, computer simulation, security, cryptography, randomized design, and other areas where producing an unpredictable result is desirable. The best RNGs for security features are hardware-based. Hardware RNGs generate random numbers as a function of current value of some physical environment attribute that is constantly changing in a manner that is practically impossible to model, thus making the guessing of a number by a malicious intruder impossible.
Random number generation is akin to rolling dice. Research shows that graphene devices can be used as hardware random number generators for secure encryption. Credit: Flickr.
The research team from Pennsylvania State University created 192 GFETs on the same chip, from Graphenea CVD graphene. The researchers noticed that due to production methods, as well as inherent local variations on the underlying substrate, electrical parameters, such as carrier mobility and doping, were not the same across all individual GFET devices. In fact, the distribution of these parameters followed Gaussian curves centered around a given mean value. The scientists used this property to create physically unclonable functions (PUF) using array of GFETs. PUFs are popular hardware security components that exploit natural variations in the physical microstructures of hardware components and their complex interactions with different stimuli (voltage, magnetic field, and light). Such a stimulus, when applied to a given microstructure, produces a unique and unpredictable result that repeats every time the same stimulus is applied to the same microstructure, but will differ between different microstructures. PUFs are often used as on-chip sources of random numbers, as well as secure keys.
To test the graphene-based solution, the scientists attacked the PUFs with machine learning algorithms with the aim of breaking security and guessing the physical parameters of GFET devices. The PUF showed resilience to the attacks that used sophisticated methods such as predictive regression models and generative adversarial neural networks.
The graphene-based PUF have the advantage of being dynamically reconfigurable with applied voltage, due to memristive properties of graphene. The graphene PUF operate with ultralow power and are scalable, stable over time, and reliable against variations in temperature and supply voltage, making them a serious candidate for applications in mass-produced real-world hardware devices.
Link to original research: Nature Electronics